Improving accuracy of long-term land-use change in coal mining areas using wavelets and Support Vector Machines

Abstract

ABSTRACTOld satellite sensors lack several quality features such as high spatial and spectral resolution. For accurate long-term change detection, improvement of the quality of old satellite images is required. In the present study, we used two wavelet-based image enhancement techniques [(discrete wavelet transform (DWT) and dual tree-complex wavelet transform (DT-CWT)] for improving the quality of Landsat 2 data of 1975 and Landsat 8 data of 2015 to study the impact of coal mining on land use change over a period of four decades. The enhanced images were subjected to land-use classification using Support Vector Machines. Land-use classification accuracy was measured using confusion matrix-based accuracy assessment. Accuracy assessment revealed that the overall classification accuracy of DWT enhanced images was 82.10% for the year 1975 and 88.46% for the year 2015. The overall classification accuracy of DT-CWT enhanced images was 85.71% for the year 1975 and 88.54% for the year 2015. The results of change detection revealed that the total areal coverage of dense vegetation increased by 65%, indicating that the rate of land degradation had slowed down due to the legislative and policy changes to promote sustainable development in coal mining after 1986.